Finance and Economics Discussion SeriesDivisions of Research & Statistics and Monetary Affairs
Federal Reserve Board, Washington, D.C.
The Bank Lending Channel of Monetary Policy and Its Effect onMortgage Lending
Lamont K. Black, Diana Hancock, and Wayne Passmore
2010-39
NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminarymaterials circulated to stimulate discussion and critical comment. The analysis and conclusions set forthare those of the authors and do not indicate concurrence by other members of the research staff or theBoard of Governors. References in publications to the Finance and Economics Discussion Series (other thanacknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.
The Bank Lending Channel of Monetary Policy and Its Effect on Mortgage Lending
Lamont Black, Diana Hancock, and Wayne Passmore*
Board of Governors of the Federal Reserve System Washington, DC 20551
May 2010
ABSTRACT The bank lending channel of monetary policy suggests that banks play a special role in the transmission of monetary policy. We look for this special role by examining the business strategies of banks as it relates to mortgage funding and mortgage lending. “Traditional banks” have a large supply of excess core deposits and specialize in information-intensive lending to borrowers (which is proxied here using mortgage lending in subprime communities), whereas “market-based banks” are funded with managed liabilities and mainly lend to relatively easy-to-evaluate borrowers. We predict that only “transition banks” operating between these business strategies are likely to increase their loan rate spreads substantially in response to monetary tightening. To fund ongoing mortgage originations, these banks must substitute from core deposits to managed liabilities, which have a large external finance premium due to these banks’ information-intensive lending. Consistent with this prediction, we find evidence of a bank lending channel only among transition banks – they significantly reduce mortgage lending in response to monetary contractions.
* The views expressed are those of the authors and do not necessarily reflect those of the Board of Governors of the Federal Reserve System or its staff. The authors would like to thank Robert Avery and Skander van den Heuvel for helpful comments as well as Lieu Hazelwood, Akeel Rangwala, Christopher Reynolds, Sarah Reynolds and Sean Wallace for their outstanding research assistance. All errors remain those of the authors.
2
I. INTRODUCTION
The bank lending channel suggests that banks play a special role in the
transmission of monetary policy. In this theory, monetary policy has an effect on banks’
cost of funds in addition to the change in the risk-free rate, leading to an additional
response in bank lending. The supply of intermediated credit therefore has a unique
response to monetary policy. To analyze the bank lending channel, we study the
response of banks to monetary policy in the context of mortgage funding and mortgage
lending. We focus on lending in subprime communities, because it is a form of
information-intensive lending which affects banks’ choices in funding sources and their
response to changes in funding costs. Our paper helps explain how mortgage loan supply
responds to monetary policy by addressing the role of banks in the transmission of
monetary policy.
We consider how two functions of banks – deposit-taking and mortgage lending –
are interrelated. In our analysis, banks’ business strategies are characterized by how they
fund mortgages and which types of mortgages they originate. “Traditional banks” use
insured retail deposits, which we refer to as core deposits, to fund information-intensive
loans. Consistent with this view, we find that traditional banks tend to extensively lend
in communities with high proportions of subprime households, where public information
about borrowers is limited. Alternatively, “market-based banks” use market debt such as
brokered deposits, which we refer to as managed liabilities, to fund loans that are easier
to evaluate. These banks tend to originate mortgages in communities of mostly prime
borrowers, where more borrower information is publicly available. We define a bank’s
core lending capacity as its ability to fund loans with core deposits; therefore, traditional
banks have a large core lending capacity whereas market-based banks have already
exhausted their core lending capacity. “Transition banks” are banks that operate between
these two business strategies, where extensive lending is done in subprime communities
but core lending capacity is sometimes exhausted.
One of the key differences between banks using these two business strategies is in
their external finance premium. Banks that must borrow uninsured funds in order to
3
make loans will have to pay an external finance premium.1 This is problematic for banks
that lend extensively in subprime communities due to the information-intensive nature of
this type of lending. Whereas prime mortgages tend to be underwritten based on a few
quantifiable factors, subprime mortgages tend to be underwritten more so on non-
quantifiable factors. This private information collected by the bank in the subprime
lending process is not available to outside investors, making the value of the balance
sheet of such lenders more difficult to evaluate. This implies that investors will charge a
larger external finance premium for lenders that extensively lend in subprime
communities.
For banks about to exhaust their core lending capacity, ongoing mortgage lending
may require the payment of an external finance premium. Deposits are limited, which
means that banks may have to switch from insured deposits to managed liabilities if they
want to continue funding new mortgages (Jayaratne and Morgan, 2000).2 For these
banks, shifting from insured retail deposits to managed liabilities creates an external
finance premium. Therefore, when banks that lend in subprime communities need to
switch from insured retail deposits to managed liabilities, their cost of funding can
quickly increase.
A significant change in banks’ cost of funds can be brought about by a change in
monetary policy, which is the basis for identifying the bank lending channel. A monetary
tightening shrinks banks’ core deposits, forcing banks to rely more on managed liabilities
and increasing their cost of funds. Our two differentiating factors, core lending capacity
and lending in subprime communities, can be combined to predict which banks are likely
to have the largest change in cost of funds due to monetary policy. Banks lending in
subprime communities will likely face a larger external finance premium if they are
required to borrow in the market and banks facing the limit of their core lending capacity
are most likely to switch from deposit funding to market funding. If the bank lending
channel exists, it most likely operates through transition banks.
1 Banks do not have to pay an external finance premium on core deposits because core deposits are insured by the government. 2 Deposits are limited because raising new deposits requires opening new branches or paying higher deposit rates.
4
Our empirical approach to identifying the bank lending channel is based on cross-
sectional variation in banks’ core lending capacity and the extent of their lending in
subprime communities. Using a model similar to the one in Kashyap and Stein (2000),
we estimate the effects of these two bank characteristics on the response of lending to
monetary policy. Whereas Kashyap and Stein focus on cross-sectional variation among
banks in their size and the liquidity of their balance sheets, we focus on banks’ core
lending capacity and lending in subprime communities as the differentiating factors in the
bank lending channel. Our approach is similar to that of Black, Hancock and Passmore
(2007) which addresses the same issue in the context of small business lending.
Mortgage lending in subprime communities is a form of “relationship lending” similar to
small business lending due to the informational opacity of the borrowers.
To set up our analysis, we stratify banks by the degree of their lending in
subprime communities and their loan-to-core deposit ratio. Our measure of subprime
lending is constructed using credit score data by census tract and the amount of each
bank’s mortgage lending by census tract. This measure is one of the paper’s
contributions, because it serves as an alternative to the HUD subprime-lender list or the
use of HMDA high-rate mortgages. Next, for each subsample of banks, we analyze the
response of mortgage lending to the federal funds rate (as a measure of monetary policy),
including interaction terms between the loan-to-core deposit ratios and the funds rate.
We identify the bank lending channel by focusing on lenders who lend in subprime
communities and likely switch into managed liabilities with an external finance premium
in response to monetary contractions. Therefore, our theoretical framework provides the
prediction that the mortgage loan growth of banks that lend in subprime communities and
have a loan-to-core deposit ratio around one should change more than that of other
lenders in response to changes in monetary policy.
Our findings are consistent with our predictions. The results indicate that banks
with core deposit funding tend to do more information-intensive mortgage lending in
subprime communities. We also find evidence that banks which extensively lend in
subprime communities and have a loan-to-core deposit ratio around one significantly
reduce lending in response to a monetary contraction. This suggests that the bank
lending channel for monetary policy applies to these banks. Additionally, these results
5
are strongest when we analyze banks’ lending to communities with a large proportion of
borrowers at the bottom of the credit score distribution. We do not find similar evidence
of a mortgage lending reduction for other banks, so we find no support for the bank
lending channel among traditional banks and market-based banks. Lastly, we consider
bank capitalization and find that, within the group of banks that have limited core lending
capacity and are originating mortgages in subprime communities, the lending response to
monetary policy is primarily among low-capitalized banks.
The remainder of our paper is organized as follows: Section II is a description of
our approach to identifying a bank lending channel. Section III describes our empirical
analysis, including a discussion of data sources, descriptive statistics, and methodology.
Section IV provides an interpretation of our results, and Section V concludes with both a
summary and policy implications for the availability of credit in subprime communities.
II. IDENTIFICATION STRATEGY
For each of the three groups of banks described above, we consider how the
spread between the loan rate and the risk-free rate would respond to a rise in the risk-free
rate. Kashyap and Stein (1995) argue that this change in spread is a measure of the bank
lending channel. When the spread remains constant (i.e. loan rates increase one-for-one
with the risk-free rate), there is no bank lending channel, but when the spread increases,
this suggests that a bank lending channel exists. The theoretical model in Black,
Hancock, and Passmore (2007) provides one explanation for why loan rates might
increase more than one-for-one with an increase in the risk-free rate due to higher loan
defaults.3 This is useful for identifying the bank lending channel, because the model
explains why the bank lending channel may exist for certain types of banks, depending
on the informational complexity of their borrowers.
In this paper, we apply the Black, Hancock, and Passmore model to test for the
presence of a bank lending channel that operates through traditional banks, market-based
banks, and transition banks. We focus on transition banks, because the impact of
3 Our identification strategy differentiates this explanation from the balance sheet channel. In the balance sheet channel, an increase in the risk-free rate causes the value of borrower collateral to fall and, perhaps, their cash flows to diminish. In response to this deterioration in credit quality, the bank would also raise loan rates more than one-for-one in response to the increase in the risk-free rate.
6
monetary policy on the loan rates of these lenders that extensively lend in subprime
communities should depend on their core lending capacity. When a transition bank
switches from core deposits to managed liabilities, the bank will pay an external finance
premium which it will pass through to its borrowers by raising loan rates. Therefore, we
predict that transition banks are the most likely to raise their interest rates (i.e., loan rate
spreads) substantially in response to a monetary tightening.
Figures 1 and 2 provide a graphical representation of the theoretical predictions
based on the model. In these figures, we differentiate between lenders with a high
proportion of lending in subprime communities and lenders with a low proportion of
lending in subprime communities. As for bank funding, we use the term “loan-to-core
deposit ratio” as the more technical version of core lending capacity.
Figure 1 shows the loan supply functions of banks according to their lending in
subprime communities and their loan-to-core deposit ratio. Lenders with a high
proportion of lending in subprime communities charge higher loan rates than lenders with
a low proportion of lending in subprime communities in order to cover their borrowers’
higher expected costs. To the degree that such banks have low loan-to-core deposit ratios
(i.e., large core lending capacities), these higher rates can be substantially mitigated
because of the economies of scope between branch deposit collection and local loan
monitoring. However, as loan-to-core deposit ratios rise, lenders with a high proportion
of subprime lending increase their loan rates more quickly because of rapidly rising costs.
In contrast, lenders that do relatively little lending in subprime communities would
increase their loan rates only slightly.
Figure 2 shows how the loan rates of each lender group respond to a shock to the
risk-free interest rate. In this figure, the vertical axis describes the multiple of the change
in the loan rate relative to a change in the risk free rate. For example, 1.2 means that a
100 basis point increase in the risk-free rate increases the loan rate 120 basis points.
Overall, lenders with a high proportion of subprime lending have a larger response to
interest rate shocks. But more importantly, this figure shows where the responses depend
most on core lending capacity. At relatively low loan-to-core deposit ratios, neither
group responds much to a shock to the risk-free interest rate because deposit funding for
new loans is readily available (securities holdings are simply decreased to free up the
7
deposits for funding loans). Similarly, at relatively high loan-to-core deposit ratios, the
responses to interest rate shocks do not depend on the banks’ loan-to-core deposit ratios
because the banks are funding loans with managed liabilities, which are readily available
for all banks.4
The responses of loan rates to an interest rate shock depend most on core lending
capacity when the loan-to-core deposit ratio is near one. As banks diminish their core
lending capacity (i.e., approach a loan-to-core deposit ratio of one from below), the
banks’ cost of supporting lending in subprime communities increases. The interaction of
these effects causes the banks to raise their loan rates all the more as the risk-free rate
rises. Close to the core lending threshold, where the loan-to-core deposit ratio equals
one, both groups of lenders show an upward tilt in how much they increase loan rates in
response to an interest rate shock. However, this upward tilt is steeper for lenders with a
high proportion of lending in subprime communities, which implies that the largest
response in loan rates should be among such banks with a loan-to-core deposit ratio close
to one.
The theory has the following predictions:
1. Lenders with a high proportion of lending in subprime communities should have a
larger core lending capacity (lower loan-to-core deposit ratio) than lenders with a
low proportion of subprime lending. Lending in subprime communities imposes
higher market premiums on these banks if they need to switch to funding
mortgages with managed liabilities.
2. Among lenders with a high proportion of lending in subprime communities, the
lenders close to exhausting their core lending capacity should raise loan rates
more sharply than the others in response to changes in monetary policy.
3. Among lenders with a large amount of core lending capacity, the amount of
deposit capacity should not affect how loan rates change in response to monetary
policy. These banks’ marginal funding does not depend on market rates.
4 In contrast to Kashyap and Stein (2000), we assume that all banks have access to uninsured sources of finance. This assumption seems valid, because, at a minimum, most banks have access to Federal Home Loan Bank advances. Therefore, rather than focusing on the liquidity of bank assets, we focus on the cost of managed liabilities.
8
4. Among lenders that have already exhausted their core lending capacity, the
amount of deposit capacity should not affect how loan rates change in response to
monetary policy. These banks’ marginal funding is already determined by market
rates and market rate increases are simply passed through to borrowers.
III. EMPIRICAL APPROACH
In this section, we describe the data sources, the grouping of banks, and the
estimation of the empirical model. All variables are constructed from 1995:Q1 to
2005:Q4, inclusive. We focus on this period prior to the dramatic increase in
securitization in 2006 and 2007 because many banks were still funding a significant
percentage of their mortgage originations using their balance sheets.
1. Data Sources
Our bank balance sheet data come from the quarterly Reports of Condition and
Income (Call Reports) for federally-insured, domestically-chartered commercial banks.
These publicly-available data provide quarterly financial statements on the banks’ assets
and liabilities. The data appendix provides details on the construction of our bank
balance sheet variables from the Call Report as well as an explanation of several filters
based on the bank data which are applied to the final sample.
We construct two measures of loans on banks’ balance sheets. To measure each
bank’s total loan portfolio in a quarter, we define total loans (LOANS) as the aggregate
book value of all loans made by the bank before the deduction of valuation reserves.
Total loans are measured on a consolidated basis (i.e., both domestic and foreign loans
are included) to provide the broadest, most comprehensive measure of each bank’s
lending. Our measure of each bank’s domestic residential mortgage lending (MORT)
includes loans secured by 1-to-4 family properties and loans extended under home equity
lines.
The main variable of interest on the liability side of the balance sheet is our
measure of core deposit funding. Core deposits (DEP) consist of the total dollar amount
of all deposits in deposit accounts of $100,000 or less. These deposits are limited to
deposits in domestic offices and include applicable accounts of transaction deposits, non-
9
transaction savings deposits, and total time deposits. Our definition of core deposits is
the same as Berlin and Mester (1999). Although the Call Report does not specifically
record the amount of insured deposits, the $100,000 threshold at the account level is a
good proxy for an account being insured by the Federal Deposit Insurance Corporation.
We also analyze the amount of securities on each bank’s balance sheet, as used by
Kashyap and Stein (2000) to measure balance sheet liquidity. Securities (SEC) include
balance sheet assets that were booked as trading assets, federal funds sold, and securities
purchased under agreement to resell. We combine these securities regardless of whether
they are classified by the bank as “held-to-maturity” or “available-for-sale.” Like total
loans, we measure securities on a consolidated basis, since either domestic or foreign
securities could potentially be liquidated by a bank to fund its lending activities.
We consider the total size of each bank and the amount of the balance sheet
funded with capital as our final balance sheet measures. Total assets (A) provide a
measure of the size of each bank’s balance sheet. Equity capital (K) includes common
stock, perpetual preferred stock, surpluses, undivided profits and capital reserves. Total
assets and equity capital, both measured on a consolidated basis, are used to calculate
bank-specific capital-to-asset ratios.
Our analysis of banking organizations is done at the top holder level to capture the
response of each organization’s overall mortgage lending to monetary policy.
Considerable evidence suggests that bank holding companies allocate capital among their
subsidiaries through the use of internal capital markets. For example, the loan growth of
banks within a bank holding company is less sensitive to cash-flow and capital (Houston
and James, 1998) as well as to monetary policy (Ashcraft, 2006) than stand-alone banks.
Given this evidence, we sum across individual bank balance sheet data (both assets and
liabilities) to construct domestic top holder balance sheet data for our analysis.5 For any
bank in a bank holding company, which is comprised of at least one bank and possibly
several banks, the bank holding company is the top holder of the banking organization.
On the other hand, a bank that is unaffiliated with any other bank is considered to be its
5 We identify the top holder of each bank using information from the National Information Center (NIC), which is the central repository containing information about all U.S. banking organizations and their domestic and foreign affiliates. Further details of the topholder construction are provided in the data appendix.
10
own top holder organization. For ease of exposition, henceforth we will refer to top
holder organizations as “banks.”
To construct measures of subprime communities, we use credit score data from
TransUnion that contains information on the credit quality of mortgage borrowers in most
U.S. census tracts. The TransUnion score of 490 is roughly comparable to a FICO score
of 600 and the TransUnion score of 600 is comparable to a FICO score of 660. We use
these data to calculate the proportion of mortgage borrowers in each census tract with a
TransUnion score below 490, which is a narrow measure of subprime borrowers, and the
proportion below 600, which is a broad measure of subprime borrowers including Alt-A.
These thresholds for subprime lending are just above and below the FICO score of 620,
which Keys et al. (2008) argue is a “rule of thumb” for subprime lending. Although we
only have the credit score distribution for 2004, we assume that the credit score
distribution for each census tract was relatively stable over the 1995 to 2005 period.
The Home Mortgage Disclosure Act (HMDA) data contain loan-level
information on the majority of mortgage lending in the United States. Under federal
requirements, almost every mortgage lender in the U.S. is required to report HMDA data
on a number of characteristics about each of their mortgage originations including the
census tract of the property and the date of origination.6 Only very small institutions,
institutions that do minimal mortgage lending, and institutions located solely in rural
areas are exempt from this requirement. For the purposes of time and product
consistency, we limit our sample to owner-occupied, conventional (non-government
guaranteed), first-lien, home-purchase mortgages, which should be relatively standard
across different lenders.7 Although the data are available beginning in 1990, the quality
of the data is not as high in the early years, so we begin our sample in 1995.
By combining the credit score data and the HMDA data, we can show the
aggregate change in subprime lending over time. Figure 3 depicts the growth of
subprime lending relative to total lending, where total lending equals our full sample of
HMDA originations and subprime lending is measured as HMDA lending within census
tracts in which the proportion of mortgage borrowers with a credit score of less than 600
6 The date stamp is only available in the confidential HMDA data. 7 The removal of government-guaranteed loans from the sample includes the removal of FHA loans.
11
is greater than 25 percent. For this figure, we include the year 2006 to illustrate that total
lending increased continuously until 2005 and then decreased in 2006. From 1995 to
2005, the number of mortgage originations increased from under 2 million to over 4.5
million and the dollar amount of mortgage originations increased from $200 billion to $1
trillion. Subprime lending followed a similar trajectory, but grew at a faster rate from
1995 to 2005 and did not decline in 2006. Therefore, from 1995 to 2006, the proportion
of mortgage lending in subprime areas increased. The proportion of mortgages by
number in low credit score tracts increased from under 0.3 to over 0.4 and the proportion
of mortgages by dollar in low credit score tracts increased from under 0.2 to over 0.3.
This shows that aggregate bank lending in subprime areas grew significantly over this
period.
Combining the TransUnion and HMDA data also allows us to generate a time-
varying measure of subprime mortgage lending for each bank. Using the HMDA data to
identify where a lender is lending (by census tract) and when a lender is lending (by
quarter), we know each bank’s geography of lending over time. This information can be
used to construct a “subprime lending intensity score” for each bank based on the credit
score distributions of the census tracts where the bank lends. We prefer this measure to
the U.S. Department of Housing and Urban Development (HUD) list of subprime lenders
because it is derived directly from our data. Additionally, our measure can be
constructed from the beginning of our sample in 1995, unlike measures of subprime
lending based on HMDA “high rate” loans, which are only available beginning in 2004.
The construction of our baseline subprime lending intensity score begins with the
proportion of mortgage borrowers with a credit score below 600 in each census tract,
which we hold fixed at 2004 levels.8 In a robustness test, we replicate the analysis using
the 490 credit score threshold. Using the credit score distribution, we then calculate an
average proportion of low credit score borrowers for each bank by weighting the
proportions in the census tracts where the bank lends by the dollar amount of the bank’s
quarterly lending in each census tract. This results in a quarterly, bank-level proportion
of mortgage borrowers with a credit score below 600. Therefore, a bank’s subprime
8 Some census tracts changed between the 1990 and 2000 Census. We use a centroid mapping of 2000 to 1990 census tracts in order to apply the appropriate credit score distribution for 1990 census tracts.
12
lending intensity score is a bank-level, weighted-average proportion of borrowers with
low credit scores at the quarterly frequency. Essentially, banks that originated mortgages
primarily in census tracts with a high proportion of subprime borrowers have a relatively
high subprime lending intensity score.
To measure changes in monetary policy that influence banks’ cost of funds, we
use the nominal federal funds rate (FFR).9 This measure has been used by a variety of
researchers (e.g., Bernanke and Blinder, 1992; Kashyap and Stein, 2000). In regard to
policy regime, Bernanke and Mihov (1998) propose that the funds rate be considered as
an appropriate measure of policy during the period we examine. In using this measure,
we associate higher values of the federal funds rate with tighter monetary policy.
Although there may be some concern about the endogeneity of the funds rate to economic
conditions, in our model the lender is responding to movements in its cost of funds. If
such changes are important, then the channel exists even if it sometimes transmits
endogenous federal funds rate changes that do not reflect the true stance of monetary
policy.
As a control for each bank’s cost of deposit funding, we use deposit market
competition, which might influence the size of the external finance premium faced by a
bank as it moves from core deposits to managed liabilities. Previous research has shown
that changes in the federal funds rate have a greater effect on bank lending in less
concentrated deposit markets (Adams and Amel, 2005; Jayaratne and Morgan, 2000).
For each bank, we construct a deposit-weighted Herfindahl-Hirshmann index (HHI) to
proxy for deposit market competition. First, branch level deposit data are used to
construct an HHI for each U.S. geographic banking market.10 We define geographic
banking markets as Metropolitan Statistical Areas (MSAs) for urban markets and as
counties for rural markets. The HHI for each market, which is calculated as the sum of
market deposit shares, increases as the concentration of banks’ deposit-taking in the
geographic market rises.11 Second, an HHI is constructed for each bank using the bank’s
9 Because our bank balance sheet data is quarterly, we use the average federal funds rate for the last month of each quarter for the value of the quarterly federal funds rate as was done in Kashyap and Stein (2000). 10 See Black, Hancock, and Passmore (2008) for details on the construction of the geographic banking market HHI. 11 For the analysis presented below, commercial bank deposits received a 100% weight and thrift deposits received a 50% weight in this calculation.
13
allocation of deposits across the markets. Each bank’s HHI is a deposit-weighted average
of the HHIs in the markets where the bank has branches.
Finally, we include information on aggregate output and regional variation to
control for loan demand. National economic conditions are measured using seasonally-
adjusted real GDP.12 To allow average economic conditions to differ across geographic
regions, we include 12 Federal Reserve-district fixed effects.
2. Lender Stratification and Descriptive Statistics
Next, we rank-order banks from the lowest to the highest subprime lending
intensity score in each quarter and stratify banks into three subprime lending groups. Our
stratification of the sample in each quarter is at the fifth and tenth percentiles of the
distribution of banks’ subprime lending intensity scores. We split the sample at these
thresholds in order to analyze the banks that lend a significant amount in subprime areas
and to compare these banks with those that do not lend very much in subprime areas.
Banks with the highest subprime lending intensity score are labeled as the “high
subprime-community lender” group and banks with the lowest subprime lending intensity
score are labeled as the “low subprime-community lender” group. The remaining banks
are labeled as the “medium subprime-community lender” group. We use the term
“subprime community,” because the score is based on lending in subprime communities.
Figure 4 shows the subprime lending intensity score thresholds in each quarter.
High subprime-community lenders have a subprime lending intensity score of more than
0.14 by the end of the sample period. This means that, on average, these lenders lend in
census tracts where more than 14 percent of the mortgage borrowers have a TransUnion
credit score of less than 600. In contrast, low subprime-community lenders have a
subprime lending intensity score of less than 0.12 by the end of the sample period. This
means that, on average, these lenders lend in census tracts where less than 12 percent of
the mortgage borrowers have a TransUnion credit score of less than 600.
Figure 5 shows the median of total assets (measured in nominal dollars) each year
for each of the three lender groups stratified by subprime lending intensity score. In
some years of the sample period, the high subprime-community lender group has greater
12 We use the chained index for real GDP in year 2000 dollars.
14
median assets than the low subprime-community lender group, and in some years it has
less. Therefore, the size of the balance sheets of banks that do a small or large proportion
of subprime lending do not appear to systematically differ, which implies that subprime
lending is not primarily associated with either small or large banks.
For each of the three lender groups, figure 6 presents a quarterly time-series plot
of the distribution of core lending capacities – measured by the loan-to-core deposit
ratios. The three panels of figure 6 show the 25th, 50th, and 75th percentiles of the
distribution of loan-to-core deposit ratios, respectively. A loan-to-core deposit ratio less
than 1 implies that the bank can fund its entire loan portfolio with core deposits.
However, when the loan-to-core deposit ratio is greater than one, the bank cannot fund all
of its loans with core deposits. These banks must use managed liabilities to fund their
loan portfolios.
Each of the panels of figure 6 shows a pattern consistent with what we would
predict. At each of the percentiles, the loan-to-core deposit ratio for the high subprime-
community lenders (represented by the light blue lines with triangles) is almost always
less than the loan-to-core deposit ratio for both the medium subprime-community lenders
(represented by the pink lines with squares) and low subprime-community lenders
(represented by the dark blue lines with diamonds). At the 75th percentile, the loan-to-
core deposit ratio for the low subprime-community lenders is also consistently greater
than the loan-to-core deposit ratio for the medium subprime-community lenders. Each of
the panels shows that loan-to-core deposit ratios have been trending upward over the
sample period, which is likely due to the falling cost of managed liabilities.13 The
median ratio for each group began below one in 1995 and trended above one over the
course of ten years. By 2005:Q4, lenders above the 75th percentile of the loan-to-core
deposit ratio for the low subprime-community lenders group had a loan-to-core deposit
ratio greater than 2.3. This means that for 25 percent of the low subprime-community
lender group in that quarter, less than a half of the loan portfolio was funded by core
deposits.
13 Membership in the Federal Home Loan Bank (FHLB) system grew substantially over our sample period, giving banks access to low-cost FHLB advances. See Frame, Hancock, and Passmore (2007).
15
To differentiate between lenders based on their core lending capacities, we group
banks according to their loan-to-core deposit ratios. As described earlier, traditional
banks have a strategy of maintaining a large core lending capacity(i.e., a low loan-to-core
deposit ratio). On the other hand, market-based banks tend to have a large loan-to-core
deposit ratio, because they have already exhausted their core deposits. We stratify banks
again by applying two thresholds to the distribution of loan-to-core deposit ratios. We
place the first threshold at 0.9, because banks with a loan-to-core deposit ratio less than
0.9 can probably continue to lend without turning to managed liabilities. The appropriate
threshold for the switching point to managed liabilities is more difficult to determine
because loan-to-core deposit ratios have been trending upward over time. Therefore, we
set the threshold at the 90th percentile of the loan-to-core deposit ratio for the full sample,
which is 1.7174 (for readability, we will sometimes refer to this threshold as 1.7). Banks
with a loan-to-core deposit ratio above this threshold have likely exhausted their core
deposits and are using managed liabilities as the marginal source of funding for loan
originations. Between the two thresholds, where loan-to-core deposit ratios fall between
0.9 and 1.7, banks are probably close to exhausting their core lending capacity and may
have to switch to managed liabilities in order to continue funding mortgages.
Table 1 presents the two-way split of banks grouped by subprime lending
intensity score and loan-to-core deposit ratio. The full panel dataset contains 84, 770
bank-quarter observations.14 In the top panel, each cell indicates the proportion of the
observations in our sample that fall within each grouping of banks. The far right column
shows the distributional splits by subprime lending at the 5 percent and 10 percent levels.
The fourth row shows the distribution of observations by loan-to-core deposit ratio.
About 25 percent of these observations are banks with a loan-to-core deposit ratio less
than 0.9 and, by construction, 10 percent of the observations are banks that have a loan-
to-core deposit ratio above 1.7. This leaves almost 65 percent of the observations as
banks with a loan-to-core deposit ratio in the middle range (top panel, fourth row, second
column) and most of these banks are included in the high subprime-community group.
Based on these percentages, it is clear that the groupings are such that we have
14 Observations in our panel dataset are comprised of a cross-section of banks being observed at a quarterly frequency.
16
maintained a large sample of banks in the group of greatest interest – high subprime-
community lenders with loan-to-core deposit ratios around one – while still
differentiating a significant proportion of other banks.
The bottom panel of table 1 shows the proportion of observations within each
grouping by subprime lending. Almost 26 percent of the high subprime-community
lenders had a loan-to-core deposit ratio below 0.9 and 9 percent of these lenders had a
loan-to-core deposit ratio above 1.7. In contrast, less than 18 percent of the low
subprime-community lenders had a loan-to-core deposit ratio below 0.9 and 21 percent of
these lenders – more than twice the proportion of high subprime-community lenders –
had a loan-to-core deposit ratio above the 1.7 threshold. Columns 1 and 3 provide
additional evidence consistent with our prediction that high subprime-community lenders
have lower loan-to-core deposit ratios than low subprime-community lenders. The
proportion of lenders with a loan-to-core deposit ratio less than 0.9 falls monotonically as
one looks down the first column from high to low subprime-community lenders, whereas
the proportion of lenders with a loan-to-core deposit ratio greater than 1.7 increases
monotonically from high to low subprime-community lenders in the third column.
3. Empirical Methodology
Our empirical approach for identifying the bank lending channel is similar to that
of Kashyap and Stein (2000). We estimate a panel regression with a cross-section of
banks and a time-series of quarters from 1995 to 2005. Let itL be the bank-level Call
Report measure of mortgage lending (MORT) at bank i at time t, let itC be the level of
the loan-to-core deposit ratio at bank i at time t ( /it it itC LOANS DEP ), and let tR be the
measure of monetary policy (with higher values of tR corresponding to tighter policy at
time t). After stratifying the data into groups based on lending to subprime communities
(high, medium, and low) and core lending capacity ( 0.9itC , 0.9 1.7174itC ,
1.7174itC ), we estimate the following relationship:
17
4 4 4 3
,1 0 0 1
12 4 4
, , 11 0 0
log logit j i t j j t j j t j t k kj j j k
k k i t i t t j t j j t j itk j j
L L R GDP TIME QUARTER
FRB HHI C TIME R GDP
where tGDP is real gross domestic product at time t, TIME is a time trend, kQUARTER
(k=1,2,3) are quarterly indicator variables, kFRB (k=1,…,12) are Federal Reserve
regional indicator variables, and itHHI is the deposit market concentration measure for
bank i at time t.
The derivatives of our regression equation depend on the cross-sectional and
time-series relationships within the data. The cross-sectional derivative , 1/it i tL C
captures the sensitivity of bank i’s mortgage lending to its core lending capacity. The
time-series derivative /it tL R measures the sensitivity of mortgage lending at bank i to
monetary policy. This derivative captures the average, direct responsiveness of bank
lending to monetary policy apart from individual bank characteristics.
We are most interested in how the sensitivity of bank lending to monetary policy
depends on each bank’s core lending capacity. Our theoretical framework implies that
the sensitivity of lending to monetary policy will depend on core lending capacity for
only a subset of banks. As in the descriptive statistics, we differentiate this group from
other banks by stratifying banks according to their subprime lending intensity score and
their loan-to-core deposit ratio.
The derivative 2, 1/it i t tL C R identifies how the sensitivity of bank lending to
monetary policy depends on each bank’s core lending capacity. Unlike the direct
measure of the lending response to monetary policy, /it tL R , this derivative exploits
both the cross-sectional and time-series aspects of the data. Suppose two high subprime-
community lenders are alike except that one is closer to exhausting its capacity to fund
loans with deposits (i.e., , 1i tC ). Now suppose that these lenders are hit by a
contractionary monetary shock. In response to that shock, our model predicts that the
lender that is closer to exhausting its core lending capacity will reduce its lending more
18
than the other lender. Therefore, for high subprime-community lenders close to
exhausting their core lending capacity, the sensitivity of their lending to monetary policy
is expected to be greater than for other high subprime-community lenders with a smaller
loan-to-core deposit ratio (i.e., 2, 1/ 0it i t tL C R ).
To test our theory, we compare within each stratified group the banks’ responses
to monetary policy given their loan-to-core deposit ratios. This is similar to Kashyap and
Stein (2000) whose strategy for identifying the bank lending channel is based on
grouping large and small banks. Their difference-in-difference approach examines how
small banks differ in response to monetary shocks based on different securities holdings.
In contrast, our approach tests the bank lending channel by examining how banks that do
a large proportion of lending in subprime communities differ from other banks in their
response to monetary policy, conditional on their loan-to-core deposit ratios.
In our framework, securities holdings are simply a place to “park” excess
deposits. A bank may draw down its securities holdings in response to a monetary shock,
but this drawdown of securities is to access core deposits and to avoid an external finance
premium, rather than to obtain cash to fund loans. This implies that a low loan-to-core
deposit ratio tends to be associated with a high securities-to-asset ratio (the balance sheet
measure used by Kashyap and Stein, 2000). As shown in table 2, loan-to-core deposit
ratios are highly negatively correlated with securities-to-asset ratios for most groups of
banks. The only exceptions are banks with high loan-to-core deposit ratios (greater than
1.7) that are also high or medium subprime-community lenders.
Bank capitalization is likely to be a contributing factor in banks’ availability of
funds and the responsiveness of their lending to monetary policy. Several authors have
found that banks with low regulatory capital are most affected by monetary contractions
(Kishan and Opiela, 2000; Kishan and Opiela, 2006). Moreover, a bank’s capitalization
seems to significantly influence the pricing of individual loans, suggesting that some
bank dependent borrowers face significant costs in switching lenders (Hubbard, Kutter,
and Palia, 2002). Bank capitalization fits easily into our analysis, as banks with low
capitalization would face an even higher external finance premium if they had to shift
from core deposits to managed liabilities. Thus, banks with a loan-to-core deposit ratio
around one would be most affected by a relatively low level of capital-to-assets. In our
19
final test, we expand our analysis by stratifying the high subprime-community lenders by
their capitalization.
We believe that our empirical approach uniquely identifies a bank loan supply
effect distinct from changes in the financial conditions of borrowers. The financial
condition of subprime borrowers may be more or less sensitive to monetary policy than
that of prime borrowers, which makes it difficult to identify a bank lending channel
where a contraction of lending reflects rising funding costs for banks and not a
deterioration of borrower balance sheets. However, we rule out this alternative
explanation because borrowers who are more prone to financial distress are not likely to
sort themselves by banks’ loan-to-core deposit ratios.
IV. EMPIRICAL RESULTS
1. Main Results
Table 3 presents estimates of the partial cross-derivatives of mortgage lending
with respect to each bank’s loan-to-core deposit ratio and monetary policy, i.e.,
2, 1/it i t tL C R , for nine subsamples of lenders stratified by subprime-community lender
group (high, medium, and low) and by core lending capacity ( 0.9itC ,
0.9 1.7174itC , 1.7174itC ). Each cell represents the degree to which a bank’s loan-
to-core deposit ratio modifies the direct effect of monetary policy on mortgage lending,
/it tL R . A negative value indicates that the lending of banks with a larger loan-to-core
deposit ratio is more sensitive to monetary policy.
The first column of table 3 reports the partial cross-derivatives of bank mortgage
lending with respect to each bank’s loan-to-core deposit ratio and the federal funds rate
for banks with a large core lending capacity ( 0.9itC ). None of these derivatives is
statistically different from zero, regardless of whether the lender is a high-, medium-, or
low- subprime-community lender. These results are consistent with our prediction that
the amount of core lending capacity should not affect how loan rates change in response
to monetary policy among lenders with a large core lending capacity.
The second column of table 3 reports the partial cross-derivatives of bank
mortgage lending with respect to each bank’s loan-to-core deposit ratio and the federal
20
funds rate for banks that have loan-to-core deposit ratios in a range around one (
0.9 1.7174itC ). These are the lenders which will most likely need to switch from
core deposit funding to managed-liability funding in order to continue funding mortgages
following a monetary contraction. In this column only the partial cross-derivative for the
high subprime-community lenders is negative and it is statistically significant at the ten
percent level. This relationship conforms to expectations: The sensitivity of lending
volume to monetary policy is greater for the high subprime-community lenders with less
core lending capacity, i.e., 2, 1/ 0it i t tL C R .
The third column of table 3 reports the partial cross-derivatives of bank mortgage
lending with respect to each bank’s loan-to-core deposit ratio and the federal funds rate
for banks that have already exhausted their core lending capacity ( 1.7174itC ).
Consistent with our theory, it is difficult to detect the presence of a bank lending channel
for these lenders, regardless of the proportion of their mortgage lending done in subprime
communities. These banks are likely already funding new mortgages with managed
liabilities, which implies that the change in their cost of funding should not depend on
their loan-to-core deposit ratio. In this column, none of the partial cross-derivatives for
the three categories of subprime lending are significant at conventional levels of
significance.
2. Robustness
As discussed in section III, the TransUnion data also have the distribution of each
census tract’s credit scores split at the TransUnion score of 490. Using this alternative
threshold, we can check the robustness of our findings to a narrower definition of
subprime lending. In this case, the subprime lending intensity score for each lender is
constructed based on the proportion of mortgage borrowers in each census tract with a
TransUnion credit score below 490 (equivalent to a FICO score of 600). For the
stratification of banks in each quarter into three subprime lending groups, we also lower
the distribution thresholds and stratify the sample at the first and fifth percentiles of the
distribution of subprime lending intensity scores. We now run an identical set of
21
regressions using our narrower definition of subprime lending along with our original
loan-to-core deposit thresholds of 0.9 and 1.7.
Table 4 shows the results for our empirical analysis based on the credit score
splits at 490. The results are qualitatively similar to the previous results shown in Table 3
for the TransUnion credit score split at 600. The negative coefficient for our main group
of interest – high subprime-community lenders with a loan-to-core deposit ratio around
one – shows that, even under this narrower definition of subprime lending, we still find
evidence for the bank lending channel. The only significant difference under the
alternative definition is the positive coefficient for the group of high subprime-
community lenders that have already exhausted their core lending capacity. This result
indicates that, among these lenders, the lending volume of banks with larger loan-to-core
deposit ratios is less sensitive to monetary policy. This might be due to some of the
lenders at the bottom of this loan-to-core deposit range transitioning from core deposit
funding to managed-liability funding.
In our final set of results, we test for the effect of bank capitalization on the bank
lending channel. We have shown that, among high subprime-community lenders with a
loan-to-core deposit ratio around one, banks with a larger loan-to-core deposit ratio have
a greater sensitivity of lending to monetary policy. It is likely that, among these banks,
those that are less well-capitalized adjust lending the most in response to monetary
contractions. To analyze this additional effect, we use a simple construction of bank
capitalization from the Call Report data, defined as capital divided by assets (K/A). We
then stratify this subset of banks (high subprime-community lender, 0.9 1.7174itC ) at
the first quartile of bank capitalization to see whether our loan-to-core deposit effect is
greatest among the least well-capitalized banks. Table 5 shows the results for both our
baseline definition of subprime lending using the credit score of 600 and the narrower
definition of subprime lending at the credit score of 490. As shown in the table, the
results are nearly identical for the alternative definitions. Only low-capital, high
subprime-community lenders (those banks that have capital-to-asset ratios in the bottom
quartile and do a large proportion of lending to subprime communities) exhibit significant
22
evidence of a bank lending channel.15 This implies that the effect of core lending
capacity on the sensitivity of loan supply among high subprime-community lenders is
primarily among low-capitalized banks.
Based on our analysis using both a broad and narrow definition of subprime
lending, the overall results appear to be slightly stronger for banks lending to the bottom
of the credit score distribution. Our analysis focuses on high subprime-community
lenders who are about to exhaust their core lending capacity. When this group is defined
as lending to borrowers with credit scores of less than 490, the results are even stronger
than when they are defined by the broader credit score of 600. In other words, subprime
mortgage loan supply is especially sensitive to monetary policy when the borrowers are
at the bottom of the credit score distribution. This can be inferred based on the
coefficients in the middle column of the first row of both tables 3 and 4, where the
coefficients are identical, but the standard error is smaller for the narrower definition of
subprime lending. Likewise, in table 5, the coefficients are nearly identical for low-
capitalized, high subprime-community lenders, yet the standard errors are smaller for this
group of lenders based on credit scores of 490. As a caveat to this inference, the group of
high subprime-community lenders under the narrow definition is slightly larger than the
baseline group due to the lender threshold being set at the fifth percentile rather than the
tenth percentile, which could narrow the standard errors. However, the smaller standard
errors are consistent with our theoretical framework, indicating that lending to lower
credit score borrowers is more information-intensive, potentially making this type of loan
supply more sensitive to monetary policy.
V. CONCLUSION
In this paper, we analyze the bank lending channel in the context of mortgage
funding and mortgage lending. The bank lending channel suggests that banks play a
special role in the transmission of monetary policy because monetary policy has an effect
on banks’ cost of funds in addition to changes in the risk-free rate. This leads to a unique
response in bank lending among certain types of banks whose funding costs change the
most in response to monetary policy.
15 Cut-offs for bank capitalization quartiles were computed on a quarter-by-quarter basis.
23
We characterize banks as “traditional banks” and “market-based banks” based on
their business strategies. Traditional banks have a large supply of excess core deposits
and specialize in information-intensive lending to borrowers, whereas market-based
banks are funded with managed liabilities and mainly lend to relatively easy-to-evaluate
borrowers. We refer to banks that operate between these two strategies as “transition
banks” because they do an extensive amount of information-intensive lending even
though their ability to continue funding mortgages with core deposits is sometimes
exhausted.
Our proxy for information-intensive lending is the amount of a bank’s mortgage
lending in subprime communities. Due to the informational opacity of such loans, banks
that extensively lend in subprime communities will likely pay a larger external finance
premium on managed liabilities. This suggests that these banks will rely more on core
deposits. However, monetary contractions can reduce banks’ supply of core deposits
and, consequently, their ability to continue funding mortgages with core deposits.
The cost of funds for transition banks will likely be the most sensitive to changes
in monetary policy due to their limited core deposits. When transition banks exhaust
their ability to fund mortgages with core deposits, they must switch to managed liabilities
and pay a large external finance premium. Therefore, it is among these banks that we
expect to find evidence for the bank lending channel of monetary policy.
Consistent with our predictions, we show that banks with an excess supply of core
deposits tend to lend more in subprime communities. We also find evidence of a bank
lending channel for banks that extensively lend in subprime communities and have a
loan-to-core deposit ratio near one. These banks significantly reduce their mortgage
lending in response to a monetary contraction. In contrast, we do not find any evidence
of a bank lending channel among traditional banks with a large core lending capacity and
among market-based banks with a large proportion of funding in managed liabilities.
This result is robust to both a broad and a narrow definition of subprime lending and
appears to be strongest among banks with low levels of capitalization. Thus, monetary
policy does seem to have a bank lending channel, which is strongest among banks that do
the most information intensive lending but have an inadequate core deposit base for
managing the costs associated with such lending.
24
DATA APPENDIX This appendix shows the details of our variable construction using the quarterly Reports of Condition and Income (Call Reports) for federally-insured, domestically-chartered commercial banks. It also provides further details on the topholder construction using Call Report and HMDA data and several filters applied to the final sample. Details of Call Report variables (RCON and RCFD) for each variable in our analysis: Total Loans: LOANS = RCFD1400. Residential Mortgages: MORT = RCON1797 + RCON1798. Core Deposits: DEP = RCON2702. Securities: The time-series for securities holdings is measured on a comparable basis through time as follows: Prior to March 2002 (SEC = RCFD8641 + RCFD3545 + RCFD1350), between March 2002 and March 2003 (SEC = RCFD8641 + RCFD3545 + RCONb987 + RCFDb989) and after March 2003 (SEC = RCFD8641 + RCFD3545 + RCFDb987 + RCFDb989). The Call Report follows generally accepted accounting principles (GAAP) by reporting securities that are “held-to-maturity” at their amortized cost and securities “available-for-sale” at their fair value. Total Assets: A = RCFD2170. Equity Capital: K = RCFD3210. Also included in equity capital are cumulative foreign currency translation adjustments less net unrealized loss on marketable equity securities. Details of Topholder Construction: Call Report filers are combined up to the topholder level. To prevent double counting of assets or liabilities when constructing top holder values, we remove banks whose Call Reports are consolidated upward. As with the Call Report data, individual HMDA filers are combined up to the topholder level to capture overall mortgage lending at the organization level. For purposes of the merge with the Call Report data, our top holders include foreign holding companies. If a
25
domestic top holder from our Call Report roll-up is part of a foreign holding company, we merge the HMDA and Call Report data at the foreign-holding-company level. Details of Sample Filters: We apply similar sample filters as those used by Kashyap and Stein (2000). Our filters are applied to Call Report data rolled up to the topholder level. First, we only keep topholders with positive assets. Second, we omit any topholders for which the ratio of mortgage loans to total loans is less than five percent. Third, we drop observations for which mortgage growth is more or less than five standard deviations from its average trend over the sample period. Lastly, we drop the observations of topholders for any quarter in which the topholders are involved in a merger.
26
REFERENCES Adams, R.M. and D.F. Amel, 2005, “The Effects of Local Banking Market Structure on
the Bank Lending Channel of Monetary Policy,” Finance and Economics Discussion Series working paper, Federal Reserve Board, 2005-16, April.
Akerlof, G., 1970, “The Market for "Lemons": Quality Uncertainty and the Market
Mechanism.” The Quarterly Journal of Economics, 84(3), pp. 488-500. Ashcraft, A., 2006, “New Evidence on the Bank Lending Channel,” Journal of Money,
Credit, and Banking, 38(3), pp. 751-775. Berlin, M. and L.J. Mester, 1999, “Deposits and Relationship Lending,” The Review of
Financial Studies, 12, Fall, pp. 579-607. Bernanke, B.S. and A. Blinder, 1992, “The Federal Funds Rate and the Channels of
Monetary Transmission,” American Economic Review, September, 83(4), pp.902-21.
Bernanke, B.S. and I. Mihov, 1998, “Measuring Monetary Policy,” Quarterly Journal of
Economics, August, 113(3), pp. 869-902. Black, L., D. Hancock, and W. Passmore, 2007, “Bank Core Deposits and the Mitigation
of Monetary Policy,” Finance and Economics Discussion Series, Federal Reserve Board of Governors, Working paper.
Demyanyk, Y. and O. Van Hemert, 2008, “Understanding the Subprime Mortgage
Crisis,” New York University, Working paper. Frame, S., D. Hancock, and W. Passmore, 2007, “Federal Home Loan Bank Advances
and Commercial Bank Portfolio Composition,” Finance and Economics Discussion Series, Federal Reserve Board of Governors, Working paper.
Gorton, G. and G. Pennacchi, 1995, “Banks and Loan Sales: Marketing Non-Marketable
Assets,” Journal of Monetary Economics, 35, pp. 389-411. Houston, J.F. and C. James, 1998, “Do Bank Internal Capital Markets Promote
Lending?” Journal of Banking and Finance, 22, pp. 899-918. Hubbard, R.G., K.N. Kuttner, and D.N. Palia 2002, “Are There Bank Effects in
Borrowers’ Costs of Funds? Evidence from a Matched Sample of Borrowers and Banks,” Journal of Business, 75, October, pp. 559-581.
Jayaratne, J. and D.P. Morgan, 2000, “Capital Market Frictions and Deposit Constraints
at Banks,” Journal of Money, Credit, and Banking, 32, pp. 74-92.
27
Kashyap, A.K. and J.C. Stein, 1995, “The Impact of Monetary Policy on Bank Balance Sheets,” Carnegie-Rochester Conference Series on Public Policy, June, pp. 151-195.
Kashyap, A.K. and J.C. Stein, 2000, “What Do a Million Observations on Banks Say
about the Transmission of Monetary Policy?” The American Economic Review, 90, June, pp. 407-428.
Keys, B., T. Mukherjee, A. Seru, and V. Vig, 2008, “Securitization and Screening:
Evidence from Subprime Mortgage Backed Securities”, working paper. Kishan, R., and T. Opiela, 2000, “Bank Size, Bank Capital, and the Bank Lending
Channel,” Journal of Money, Credit, and Banking, 32, pp. 121-41. Kishan, R., and T. Opiela, 2006, “Bank Capital and Loan Asymmetry in the
Transmission of Monetary Policy,” Journal of Banking and Finance, 30, pp. 259-85.
Nakamura, L., 1993, “Commercial Bank Information: Implications for the Structure of
Banking.” In: Klausner, Michael, and Lawrence J. White, eds.: Structural Change in Banking, pp. 131-160.
Quercia, R., M. Stegman, and W. Davis, 2007, “The Impact of Predatory Loan Terms on
Subprime Foreclosures: The Special Case of Prepayment Penalties and Balloon Payments,” Housing Policy Debate 18, pp. 311-345.
Pennington-Cross, A., 2003, “Credit History and the Performance of Prime and
Nonprime Mortgages,” Journal of Real Estate Finance and Economics 27, pp. 279-301.
28
* Assumptions based on the theory developed in Black, Hancock, and Passmore (2007) using a lognormal distribution of returns for the borrowers: For the lenders with a high proportion of subprime lending and a loan-to-core deposit ratio less than 100 percent, ρ = 0.075, δ = 0.80, φ = 0.00, σ = 2.25, μ = 2.25. For the lenders with a high proportion of subprime lending and a loan-to-core deposit ratio greater than 100 percent, the same parameters hold, except φ = 0.15. For lenders with a low proportion of subprime lending and a loan-to-core deposit ratio less than 100 percent, ρ = 0.010, δ = 0.95, φ = 0.00, σ = 2.25, μ = 2.25. For lenders with a low proportion of subprime lending and a loan-to-core deposit ratio greater than 100 percent, the same parameters hold, except that φ = 0.05. For both types of lenders, loan demand is assumed to be linear (LD= c -β*rL ). However, equilibrium loan rates can also be derived (as well as similar charts) for LD= (α*q -rL)/β, which is a non-linear demand function that is dependent on the payoff and probability of success of the firm’s project.
Figure 1
Core Lending Capacity and Its Effect on Loan Rates
Lenders with a High Proportion of Subprime Lending
h Lenders with a Low Proportion of Subprime Lending
29
* Assumptions based on the theory developed in Black, Hancock, and Passmore (2007) using a lognormal distribution of returns for the borrowers: For the lenders with a high proportion of subprime lending and a loan-to-core deposit ratio less than 100 percent, ρ = 0.075, φ = 0.80, σ = 2.25, μ = 2.25. For the lenders with a high proportion of subprime lending and a loan-to-core deposit ratio greater than 100 percent, the same parameters hold, except φ = 0.15. For the lenders with a low proportion of subprime lending and a loan-to-core deposit ratio less than 100 percent, ρ = 0.010, δ = 0.95, φ = 0.00, σ = 2.25, μ = 2.25. For the lenders with a high proportion of subprime lending, β = -0.01, and for the lenders with a low proportion of subprime lending, β = -2.00. For the lenders with a low proportion of subprime lending and a loan-to-core deposit ratio greater than 100 percent, the same parameters hold, except that φ = 0.05. For both types of lenders, loan demand is assumed to be linear (LD= c - β*rL ). However, equilibrium loan rates can also be derived (as well as similar charts) for LD = (α*q -rL)/β, which is a non-linear demand function that is dependent on the payoff and probability of success of the firm’s project.
Figure 2
Effect of Core Lending Capacity on the Spread
Lenders with a High Proportion of Subprime Lending
h Lenders with a Low Proportion of Subprime Lending
30
Figure 3 Aggregate Mortgage Lending in Low Credit Score Tracts, 1995 to 2006
(600)
Number of Mortgage Originations by Year
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
N u m
b e r ( i n
t h o u s a n d s )
Total
Low Credit Score Proportion Less Than or Equal to 0.25
Low Credit Score Proportion More Than 0.25
Proportion by Year
0.15
0.2
0.25
0.3
0.35
0.4
0.45
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
P r o p o r t i o n
Proportion of Mortgages by Number in Low Credit Score Tracts
Proportion of Mortgages by Dollar in Low Credit Score Tracts
Dollar Amount of Mortgage Originations by Year
0
200
400
600
800
1000
1200
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
D o l l a
r A m
o u n t s (
i n m
i l l i o n s )
Total
Low Credit Score Proportion Less Than or Equal to 0.25
Low Credit Score Proportion More Than 0.25
31
Figure 4 Subprime Intensity Score Cut-offs
(600)
0.08
0.09
0.1
0.11
0.12
0.13
0.14
0.15
1995
Q1
1995
Q3
1996
Q1
1996
Q3
1997
Q1
1997
Q3
1998
Q1
1998
Q3
1999
Q1
1999
Q3
2000
Q1
2000
Q3
2001
Q1
2001
Q3
2002
Q1
2002
Q3
2003
Q1
2003
Q3
2004
Q1
2004
Q3
2005
Q1
2005
Q3
Date
Su
bp
rim
e L
end
ing
Inte
nsi
ty S
core
High Subprime-Community Lender Cut-offs (10%)
Low Subprime-Community Lender Cut-offs (5%)
32
Figure 5 Median of Total Assets by Subprime-Community Lender Group, 1995:Q1 – 2005:Q1
(600)
0
50000
100000
150000
200000
250000
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Date
To
tal A
sset
s (
in t
ho
usan
ds
of
US
$)
Low Subprime-Community Lenders Medium Subprime-Community Lenders High Subprime-Community Lenders
33
Figure 6 Loan-to-Core Deposit Ratios by Subprime-Community Lender Group
(600)
25th Percentiles
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1995
Q1
1995
Q3
1996
Q1
1996
Q3
1997
Q1
1997
Q3
1998
Q1
1998
Q3
1999
Q1
1999
Q3
2000
Q1
2000
Q3
2001
Q1
2001
Q3
2002
Q1
2002
Q3
2003
Q1
2003
Q3
2004
Q1
2004
Q3
2005
Q1
2005
Q3
Date
Lo
an
-to
-Co
re D
ep
osit
Rati
o
Low Subprime-Community Lenders
Medium Subprime-Community Lenders
High Subprime-Community Lenders
Medians
0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
1995
Q1
1995
Q3
1996
Q1
1996
Q3
1997
Q1
1997
Q3
1998
Q1
1998
Q3
1999
Q1
1999
Q3
2000
Q1
2000
Q3
2001
Q1
2001
Q3
2002
Q1
2002
Q3
2003
Q1
2003
Q3
2004
Q1
2004
Q3
2005
Q1
2005
Q3
Date
Lo
an
-to
-Co
re D
ep
osit
Rati
o
Low Subprime-Community Lenders
Medium Subprime-Community Lenders
High Subprime-Community Lenders
75th Percentiles
0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
2.1
2.3
2.5
1995
Q1
1995
Q3
1996
Q1
1996
Q3
1997
Q1
1997
Q3
1998
Q1
1998
Q3
1999
Q1
1999
Q3
2000
Q1
2000
Q3
2001
Q1
2001
Q3
2002
Q1
2002
Q3
2003
Q1
2003
Q3
2004
Q1
2004
Q3
2005
Q1
2005
Q3
Date
Lo
an
-to
-Co
re D
ep
osit
Rati
o
Low Subprime-Community Lenders
Medium Subprime-Community Lenders
High Subprime-Community Lenders
34
Table 1 How Many Banks Are Core Lending Capacity Constrained?
Proportion of Bank-Quarter Observations by Subprime-Community Lender Group and by Range of Loan-to-Core Deposit Ratio (Subprime Lending Intensity Score Cut-offs at 5% and 10%)
(600)
Proportional Basis: Subprime-Community Lender Group
Loan-to-Core Deposit Ratio Range
Less than .9
.9 to 1.7174
Greater than 1.7174
Total
Proportion of All Observations
High Subprime 0.234 0.586 0.081 0.900 Medium Subprime 0.009 0.032 0.009 0.050 Low Subprime 0.009 0.030 0.010 0.050 Total 0.252 0.648 0.100 1.000 Proportion of Each Group
High Subprime 0.259 0.651 0.090 1.000 Medium Subprime 0.183 0.639 0.178 1.000 Low Subprime
0.178
0.611 0.211
1.000
35
Table 2 The Relationship between Loan-to-Core Deposit Ratios and Securities-to-Assets Ratios
Pearson Correlation Coefficients (P-values in parentheses)
(600)
Subprime-Community Lender Group
Loan-to-Core Deposit Ratio Range
Less than 0.9
0.9 to 1.7174
Greater than 1.7174
High Subprime
-0.724
(<0.001)
-0.435
(<0.001)
0.025
(0.036)
Medium Subprime
-0.650
(<0.001)
-0.346
(<0.001)
0.213
(<0.001)
Low Subprime
-0.605
(<0.001)
-0.349
(<0.001)
-0.028 (0.401)
36
Table 3 Partial Derivatives of Bank Mortgage Lending Growth
With Respect to the Monetary Policy (R) and Core Lending Capacity (C) (P-Values in parentheses)
(Subprime Thresholds Defined at Credit Score of 600)
Subprime-Community Lender Group
Loan-to-Core Deposit Ratio Range
Less than 0.9
0.9 to 1.7174
Greater than 1.7174
High Subprime (above 10%)
0.023
(0.215)
-0.020 (0.009)
-0.001 (0.854)
Medium Subprime
0.062
(0.572)
-0.033
(0.344)
-0.008 (0.483)
Low Subprime (below 5%)
0.072
(0.592)
-0.027 (0.368)
-0.010 (0.369)
Difference (High minus Low)
-0.048 (0.713)
0.008
(0.801)
0.009
(0.523)
37
Table 4 Partial Derivatives of Bank Mortgage Lending Growth
With Respect to the Monetary Policy (R) and Core Lending Capacity (C) (P-Values in parentheses)
(Subprime Thresholds Defined at Credit Score of 490)
Subprime-Community Lender Group
Loan-to-Core Deposit Ratio Range
Less than 0.9
0.9 to 1.7174
Greater than 1.7174
High Subprime (above 5%)
0.026
(0.166)
-0.020 (0.006)
0.007
(0.022)
Medium Subprime
0.181
(0.241)
-0.027 (0.448)
0.013
(0.414)
Low Subprime (below 1%)
-0.237 (0.407)
0.022
(0.804)
-0.025 (0.489)
Difference (High minus Low)
0.263
(0.279)
-0.042 (0.617)
0.031
(0.332)
38
Table 5 Does Bank Capitalization Matter for Banks That Are Core Lending Capacity Constrained?
Partial Derivatives of Bank Mortgage Lending Growth With Respect to the Monetary Policy (R) and Loan-to-Core Deposit Ratio (C) For Banks with a Loan-to-Core Deposit Ratio around One (0.9 < C < 1.7174)
(P-Values in parentheses)
Subprime-Community Lender Group
Level of Capitalization
Capital-to-Asset Ratio in the Bottom Quartile
Capital-to-Asset Ratio in the Top Three Quartiles
High Subprime (based on credit score of 600)
High Subprime (based on credit score of 490)
-0.048 (0.0039)
-0.049 (0.0021)
-0.008 (0.3268)
-0.008 (0.2918)
Top Related